CN110110766A - A kind of online character analysis method and device based on motion planning controlling feature - Google Patents
A kind of online character analysis method and device based on motion planning controlling feature Download PDFInfo
- Publication number
- CN110110766A CN110110766A CN201910329694.6A CN201910329694A CN110110766A CN 110110766 A CN110110766 A CN 110110766A CN 201910329694 A CN201910329694 A CN 201910329694A CN 110110766 A CN110110766 A CN 110110766A
- Authority
- CN
- China
- Prior art keywords
- feature
- motion planning
- control
- adjustment
- sliding
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F11/00—Error detection; Error correction; Monitoring
- G06F11/30—Monitoring
- G06F11/34—Recording or statistical evaluation of computer activity, e.g. of down time, of input/output operation ; Recording or statistical evaluation of user activity, e.g. usability assessment
- G06F11/3438—Recording or statistical evaluation of computer activity, e.g. of down time, of input/output operation ; Recording or statistical evaluation of user activity, e.g. usability assessment monitoring of user actions
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/23—Clustering techniques
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Data Mining & Analysis (AREA)
- General Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Artificial Intelligence (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Evolutionary Biology (AREA)
- Evolutionary Computation (AREA)
- Quality & Reliability (AREA)
- Computer Hardware Design (AREA)
- User Interface Of Digital Computer (AREA)
Abstract
A kind of online character analysis method and device based on motion planning controlling feature, analysis method include: Step 1: obtaining and recording the motion planning control data of active user；Step 2: carrying out data prediction to motion planning control data, the sample action of user is obtained；Step 3: the motion planning mode in analysis sample action, extracts the motion planning controlling feature in sample action；Step 4: carrying out character analysis to motion planning controlling feature by the personality model pre-set, wherein the personality model based on motion planning controlling feature is the different characters property sort device or recurrence device obtained by the training of several sample actions.Device includes data capture unit, pretreatment unit, feature extraction unit, character analysis unit.Online character analysis method data source procurement cost of the invention is low, and can effectively analyze user's personality.
Description
Technical field
The present invention relates to computer and mobile network user information Perception analysis fields, and in particular to one kind is based on movement rule
The online character analysis method and device for drawing controlling feature, analyzes user by the use to computer or intelligent mobile terminal
Personality.
Background technique
With social informatization, the propulsion of networking spring tide, to the perception of user information in computer and mobile network
Analysis becomes more and more important, in the virtual interaction systems such as e-commerce, electronic instruction, virtual team, system there is an urgent need to
Understand the personality of system user, to formulate efficient system scheme, provides personalized user service.For example, in electronics quotient
In the network virtualizations economic activity such as business, finance and money management, businessman is highly desirable can adequately to understand client as far as possible, determine target
The personality of client, businessman can provide targetedly commodity or service for client, to improve the success rate of business activity.
The character analysis method that researcher proposes now is based primarily upon social networks, mobile phone use habit, audio, video data
And physiological signal.Based on social networks post content, Analysis of Network Information user's personality such as thumb up and need to obtain current system
The social networks account of user in e-commerce system, is usually not easy to obtain network of relation data, however in real system
Know the social network account of current E-commerce system user, it is difficult to obtain related data.Property based on mobile phone use habit
Case analysis method needs to obtain a large number of users information, including call short message record, bluetooth, position, microphone, sensor etc., has
A little relevant data of content can be related to individual privacy.And the method based on audio, video data and physiological signal needs specific information
Equipment, such as camera are acquired, EEG signal acquires equipment etc., is not suitable for active computer network environment.
Summary of the invention
It is an object of the invention to be directed to above-mentioned the problems of the prior art, provide a kind of based on motion planning controlling feature
Online character analysis method and device, data source procurement cost is low and can effectively analyze user's personality.
To achieve the goals above, the present invention is based on the online character analysis methods of motion planning controlling feature, including with
Lower step:
Step 1: obtaining and recording the motion planning control data of active user；
Step 2: carrying out data prediction to motion planning control data, the sample action of user is obtained；
Step 3: the motion planning mode in analysis sample action, extracts the motion planning controlling feature in sample action；
Step 4: carrying out character analysis to motion planning controlling feature by the personality model pre-set, wherein base
In the personality model of motion planning controlling feature be by several sample actions obtained different characters property sort device of training or
Return device.
The motion planning control number of active user is obtained and recorded by computer, mobile terminal or wearable device
According to；
In the step one, when obtaining motion planning control data, user is based on current motion state and target
The gap of motion state is planned and adjusts current motion state to reach target state；
The motion planning mode for obtaining motion planning control data includes mouse running fix target motion planning mode, key
Disk inputs predetermined sequence motion planning mode, touch screen input prearranged gesture motion planning mode and everyday actions motion planning
Mode；
The mouse running fix target motion planning mode are as follows: user is using mouse control computer screen glazing
Mark it is mobile and make it position when, present cursor position and target on the computer comparison screen of view-based access control model perception
Location information, it is dynamic change that operation mouse, which moves closer to the cursor controlled on screen and reach the process of target position,
It plans and adjusts comprising movement routine, movement speed plans the mobile adjustment and control for adjustment, moving closer to object procedure；
The keyboard inputs predetermined sequence motion planning mode are as follows: user is to complete predetermined keystroke sequence using keyboard defeated
When the target entered, based on the mobile keystroke of the relative position information of finger and keyboard control finger, what reflection finger movement was coordinated is hit
Key rhythm；
The touch screen inputs prearranged gesture motion planning mode are as follows: user is defeated using touch screen completion prearranged gesture
When the target entered, the display information change based on intelligent mobile terminal screen, control finger sliding is to complete corresponding gesture input
Process be dynamic change, comprising sliding pressure adjust with control, sliding speed adjustment with control, sliding path adjustment and control
System；The prearranged gesture includes pull-up, drop-down, left cunning, right cunning, zoom-in and zoom-out；
The everyday actions motion planning mode are as follows: user completes daily in wearing mobile terminal or wearable device
When movement, habit is controlled by the everyday actions that the data that motion sensor records obtain user.
Sample action includes mouse action sample, keyboard key stroke sample, touch screen slides sample, motion sensor records
Everyday actions sample；Motion planning controlling feature includes that mouse running fix target motion planning controlling feature, keyboard input are pre-
Sequencing column motion planning controlling feature, touch screen input prearranged gesture motion planning controlling feature and everyday actions motion planning control
Feature processed；
Interactive device, user when motion planning control data pass through user's operation computer or intelligent mobile terminal are worn
Mobile terminal and the motion sensor measurement of wearable device obtain；Wherein, user's operation computer or intelligent mobile are whole
Interactive device when end includes mouse, keyboard, touch screen, and interbehavior data include mouse action data, keyboard key stroke data
Data are slided with touch screen, everyday actions data are recorded by the motion sensor of mobile terminal or wearable device.
The mouse running fix target motion planning controlling feature includes the planning of characterization mouse movement routine and adjustment
Feature, the planning of characterization mouse movement speed and the feature of adjustment, characterization mouse move closer to the mobile adjustment of object procedure with
The feature of control；
1) planning of characterization mouse movement routine and the feature of adjustment include:
During mobile mouse, when cursor position deviates central axes farther out, user, which adjusts move angle, makes light for description
Mark return central axes when coordinate to target location coordinate distance move angle-distance coordination feature；
Traveling time and present cursor position is described to assist to traveling time-distance of relationship between the distance of target position
Adjust feature；
Mobile curvature-the time coordination for describing mobile bending degree dynamic adjustment and traveling time relationship between the two is special
Sign；
The mobile curvature statistic of mobile bending degree globality variation is described；
The mobile deviation central axes statistic of the mobile offset central axes globality variation of description；
The moving displacement of mobile efficiency and the ratio of distance are described；
2) planning of characterization mouse movement speed and the feature of adjustment include:
The mobile dynamically distance of adjustment and present cursor position to target position that accelerates and slow down of description is closed between the two
The movement speed of system-distance coordinates feature；
Description is mobile to be accelerated and slows down movement speed-time coordination of dynamic adjustment and traveling time relationship between the two
Feature；
The movement speed statistic of movement speed globality variation is described；
Movement speed is described with the translational acceleration statistic of time change；
3) feature for mobile adjustment and the control that characterization mouse moves closer to object procedure includes:
The ratio close to target traveling time and whole section of traveling time close to target adjustment time span is described；
Describe to change close to movement speed globality in the moving process of target close to target velocity statistic；
Description close in the moving process of target mobile curvature globality variation close to target curvature statistic；
The ratio close to target moving displacement and distance of the mobile efficiency close in the moving process of target is described；
Residence time of the mobile finish time of mouse task switching efficiency to the click moment between the two is described；
Whether moving for description mistake adjustment is back adjusted more than target to reach target.
The keyboard input predetermined sequence motion planning controlling feature includes the keystroke section that characterization reflection finger movement is coordinated
The feature played；The feature for the keystroke rhythm that the characterization reflection finger movement is coordinated includes the key-efficient for describing each key
Keystroke residence time statistic, the keystroke flight time statistic of key-efficient of the adjacent key twice of description, description it is given
Two letters or the corresponding key pair of number key-efficient double character group keystroke flight time statistics.
The touch screen input prearranged gesture motion planning controlling feature includes the spy for characterizing sliding pressure adjustment and control
Sign, the adjustment of characterization sliding speed and the feature of control, the feature of characterization sliding path adjustment and control；
1) feature of the characterization sliding pressure adjustment and control includes:
Describe sliding pressure dynamic adjustment with sliding distance between the two relationship sliding pressure-distance coordination feature；
Sliding pressure-time coordination feature of sliding pressure dynamic adjustment and sliding time relationship between the two is described；
The sliding pressure statistic of sliding pressure globality variation is described；
2) feature of the characterization sliding speed adjustment and control includes:
Describe sliding speed dynamic adjustment with sliding distance between the two relationship sliding speed-distance coordination feature；
Sliding speed-time coordination feature of sliding speed dynamic adjustment and sliding time relationship between the two is described；
The sliding speed statistic of sliding speed globality variation is described；
The sliding acceleration statistic of description sliding acceleration globality variation；
3) feature of the characterization sliding path adjustment and control includes:
The slip distance of sliding length is described；
Describe glide direction dynamic adjustment with sliding distance between the two relationship glide direction-distance coordination feature；
Glide direction-time coordination feature of glide direction dynamic adjustment and sliding time relationship between the two is described；
Sliding curvature-time coordination of description sliding bending degree dynamic adjustment and sliding time relationship between the two is special
Sign；
The sliding curvature statistic of description sliding bending degree.
Everyday actions motion planning controlling feature includes the feature for characterizing the travelling control habit in user's walking process.
Characterizing the feature that the travelling control in user's walking process is accustomed to includes:
It describes user to walk the feature of rhythm, including walking cycle, acceleration of averagely walking, the average speed of travel, average
Step-length；When describing user's walking, the body bobbing feature stopped over body bobbing rule during foot-up；One row of user is described
Walking acceleration-the travel time for walking the adjustment of the dynamic of acceleration in the period and travel time relationship between the two coordinates feature.
The step four is carried out as follows foundation based on the personality model of motion planning controlling feature:
4-1, several sample actions are obtained to several motion plannings control data progress data prediction；It analyzes several
Motion planning mode in a sample action, and spy is controlled for the motion planning in the motion planning schema extraction sample action
Sign；
4-2, dimensionality reduction is carried out to the motion planning controlling feature by feature selecting algorithm；
4-4, classified using sorting algorithm to the motion planning controlling feature after dimensionality reduction, or use regression algorithm pair
Motion planning controlling feature after dimensionality reduction is returned, and the personality mould that different characters characteristic corresponds to motion planning controlling feature is obtained
Type.
The present invention is based on the online character analysis devices of motion planning controlling feature:
Including the data capture unit for obtaining and recording active user's motion planning control data, for record
Motion planning control data carry out data prediction and obtain the pretreatment unit of user action sample, for analyzing in sample action
Motion planning mode and for the feature extraction of the motion planning controlling feature in the motion planning schema extraction sample action
Unit, and the character analysis for carrying out character analysis to motion planning controlling feature according to the personality model pre-set
Unit.
Preferably, if the pretreatment unit can control several motion plannings data progress data prediction and obtain
Dry sample action, feature extraction unit can analyze the fortune in the motion planning schema extraction sample action in several sample actions
Dynamic planning control feature carries out dimensionality reduction to motion planning controlling feature by Feature Dimension Reduction unit, is adopted by model learning unit
Classified with sorting algorithm to the motion planning controlling feature after dimensionality reduction, or the movement after dimensionality reduction is advised using regression algorithm
It draws controlling feature to be returned, obtains the personality model that different characters characteristic corresponds to motion planning controlling feature；
The feature extraction unit includes that mouse running fix target motion planning controlling feature extraction unit, keyboard are defeated
Enter predetermined sequence motion planning controlling feature extraction unit, touch screen input prearranged gesture motion planning controlling feature extraction unit
And everyday actions motion planning controlling feature extraction unit；Wherein, mouse running fix target motion planning control
Feature, characterization mouse movement speed planning and tune of the feature extraction unit to extract the planning of characterization mouse movement routine and adjustment
Whole feature, characterization mouse move closer to the feature of the mobile adjustment and control of object procedure；The keyboard inputs pre- sequencing
The feature for the keystroke rhythm that the characterization reflection finger movement of column motion planning controlling feature extracting unit used for extracting is coordinated；Described
Touch screen inputs the spy of prearranged gesture motion planning controlling feature extracting unit used for extracting characterization sliding pressure adjustment and control
Sign, the adjustment of characterization sliding speed and the feature of control, the feature of characterization sliding path adjustment and control；The everyday actions fortune
Feature of the dynamic planning control feature extraction unit to extract the travelling control habit in characterization user's walking process；
The model learning unit includes projection subelement, classification subelement, returns subelement and storing sub-units；Projection
Subelement is to use mapping function by the motion control Projection Character after dimensionality reduction to higher dimensional space, and subelement of classifying is to be based on
Higher dimensional space motion control feature training classification method after projection returns subelement to based on the higher dimensional space fortune after projection
Dynamic controlling feature training homing method, storing sub-units to training result is stored as different characters characteristic disaggregated model or
Regression model.
Compared with prior art, the present invention is with following the utility model has the advantages that since people is whole using computer, intelligent mobile
The motion planning control data of consecutive variations and dynamic adjustment can be generated during end, which includes friendship
Mutual behavioral data and motion sensor data, wherein interbehavior data include mouse running fix target data, keyboard input
Predetermined sequence data and touch screen input prearranged gesture data, and motion sensor data includes acceleration transducer and gyroscope note
The everyday actions data of record.A series of environment sensing-fortune that motion planning control data feature consecutive variations and dynamic adjusts
Dynamic planning control process, the process include information Perception and information processing, decision thinking and motor imagination, move and recognize with brain
Know that the activity of corresponding functional structure is closely related, and personality is also influenced by identical cerebral function structure, therefore the present invention
The motion control characteristic model for clustering or returning various dimensions by machine learning method analyzes user's personality.It is of the invention
Linear lattice analysis method data source procurement cost is low, and can effectively analyze user's personality, is guaranteed by related device
The realization and application of method.
Detailed description of the invention
Character analysis method flow diagram of the Fig. 1 based on mouse running fix target signature；
Character analysis apparatus structure block diagram of the Fig. 2 based on mouse running fix target signature；
Personality model training method flow chart of the Fig. 3 based on mouse running fix target signature；
Personality model training apparatus structural block diagram of the Fig. 4 based on mouse running fix target signature.
Specific embodiment
The present invention is described in further detail with reference to the accompanying drawings and embodiments.
Referring to Fig. 1, the present invention is based on the character analysis method of mouse running fix target signature the following steps are included:
Step 101: user obtains and records the mouse action data of user when interacting with a computer using mouse；
Available data are as shown in table 1:
The available mouse action data of table 1
| Action type | Data format |
| Mouse is mobile | { mouse is mobile, x coordinate y-coordinate, timestamp } |
| Mouse is clicked | { mouse is clicked, x coordinate y-coordinate, timestamp } |
Step 102: data prediction being carried out to the mouse action data of record and obtains the mouse action sample of user.
The data prediction may include data resampling, data smoothing.
Wherein, the mouse action data that data resampling is used to will acquire are spaced to schedule carries out resampling, this
The predetermined time interval of inventive embodiments is 8ms.Data smoothing is to be carried out the data after resampling using data smoothing algorithm
Denoising disposal, data smoothing algorithm used in the embodiment of the present invention are sliding averaging method.Those skilled in the art can be with logarithm
It is separately selected according to the predetermined time interval and data smoothing algorithm of resampling, data resampling steps can also be skipped directly to obtaining
The mouse action data taken carry out Denoising disposal using data smoothing algorithm, do not influence the realization of the method for the present invention.
Step 103: for mouse action sample, cursor is mobile to approach simultaneously on analysis operation mouse control computer screen
Make it dynamic change during position: the planning of user's dynamic change movement routine is planned with adjustment, movement speed
With the mobile adjustment and control for adjusting, moving closer to object procedure；It extracts mouse running fix target signature and describes movement rule
Draw control model.
In an embodiment of the present invention, the example of reference table 2-1,2-2,2-3, table 2-1,2-2,2-3 image are shown
The particular content of mouse running fix target signature, example include the feature for characterizing the planning of mouse movement routine and adjustment, characterization
The feature of mouse movement speed planning and adjustment, and characterization mouse move closer to the spy of the mobile adjustment and control of object procedure
Sign.
Table 2-1 characterizes the feature of mouse movement routine planning and adjustment
In table 2-1, traveling time-distance coordinates the mobile mouse of characteristic measure user and completes different mobile phase need
The traveling time to be spent describes the relationship of traveling time at a distance from present cursor position to target position between the two.
Specifically, when calculating, the exemplary different mobile phase of table 2-1 is respectively current position coordinates to target position
The distance of coordinate is 25%, 50%, 75%, the 100% of whole section of moving distance.Curvature-time coordination characteristic measure is moved to exist
User's different traveling time stage, the average bending degree of motion track, specifically, when calculating, the exemplary difference of table 2-1
The traveling time stage is (0~0.25) T, (0.25~0.5) T, (0.5~0.75) T, (0.75~1) T.
In practical applications, those skilled in the art can choose other mobile phases, the traveling time stage is distinguished
It calculates traveling time-distance and coordinates feature and mobile curvature-time coordination feature, do not influence the realization of technical solution of the present invention.
Table 2-2 characterizes the feature of mouse movement routine planning and adjustment
In table 2-2, the speed that movement speed-distance coordinates the mobile mouse of characteristic measure user is raised and lowered into not
With velocity amplitude when corresponding mobile phase, describe the mobile dynamic adjustment that accelerates and slow down of user and arrived with present cursor position
The relationship of the distance of target position between the two.In the speed of the mobile mouse of movement speed-time coordination characteristic measure user
The traveling time for rising and needing to spend when dropping to different velocity amplitudes, describe user it is mobile accelerate and slow down dynamic adjustment with
The relationship of traveling time between the two, specifically, the exemplary different velocity amplitude of table 2-2 is current movement speed when calculating
25%, 50%, 75%, 100% and current movement speed for rising to whole section of mobile movement speed maximum value drop to whole section
25%, 50%, the 75% of mobile movement speed maximum value.In practical applications, those skilled in the art can choose other
Velocity amplitude coordinate feature and movement speed-time coordination feature to calculate separately movement speed-distance, do not influence skill of the present invention
The realization of art scheme.
Table 2-3 characterization mouse moves closer to the feature of the mobile adjustment and control of object procedure
Mouse running fix target motion planning controlling feature in table is only signal of illustrating, and those skilled in the art can
To select the feature of any one in table or any combination to carry out character analysis, the selection of feature does not influence reality of the invention
It is existing.
Step 104: according to the personality model based on mouse running fix target signature of training in advance, being based on step 103
Obtained mouse running fix Target Signature Analysis user's personality；Training method process is as shown in Figure 3.
The personality measure that the embodiment of the present invention is chosen includes 4 character dimensions, and every character dimension includes two kinds of property
Lattice are biased to, and are respectively corresponded are as follows: export-oriented-internally-oriented, Intuition-modality of sensation, thoughtful type-emotion type, judgement type-sensing type.Its
In, export-oriented-internally-oriented measure user is more concerned about the external world to the attitude of attention and energy, export-oriented user, it is easier to
With the external world generate contact, it is intended to by energy be directed to other people and actively from interacted with other people and participation activity in obtain
Energy, internally-oriented user are more concerned about inner world, are more willing to individually or do things together with less people, it is intended to draw energy
It leads inner world and energy is obtained by the thinking of inner world.Intuition-modality of sensation personality dimension measure user is being believed
Information Perception mode different in perception is ceased, the user for measuring different deviations is more likely to the information type of concern, intuition
Type user tend to pay close attention to current information in contain mode, meaning and future a possibility that, it is intended to pay close attention to different information it
Between connection, quickly form the big figure of a width and then focus on the fact, modality of sensation user mainly passes through five kinds of sensory reception's information, and
Pay attention to the fact that currently occur and details, it is intended to the big figure of a width is then formed since collecting the actual facts and details.Thoughtful type-feelings
Sense type personality dimension measure user decision thinking mode different during decision thinking is more inclined when measuring different user decision
To in the decision criteria of concern, thoughtful type user more payes attention to logic analysis, objective standard and impersonal when making a decision
The fact, emotion type user more pay attention to the harmony of personal values and environment in decision.Judgement type-sensing type measure user exists
Attitude when interacting with the external world judges that type user tends to more planned, coherent life style, and preference is very fast
Make a policy, sensing type user tends to more flexible life style, and preference keeps options open.In analysis user's property
When lattice, the present embodiment uses machine learning classification algorithm, as support vector machines infers that user is inclined in the personality of every character dimension
To.
Corresponding with the above-mentioned character analysis method based on mouse running fix target signature, the present invention also provides one kind
Based on the character analysis device of mouse running fix target signature, referring to fig. 2, the present embodiment may include following unit:
Data capture unit 201 is obtained and is recorded to user data, in embodiments of the present invention, to make in user
When being interacted with a computer with mouse, the mouse action data of user are obtained and record, available data are shown in Table 1.
Pretreatment unit 202 obtains the mouse action sample of user to carry out data prediction to record mouse action data
This.
Wherein, the pretreatment unit 202, can specifically include:
Data resampling sub-units, the mouse action data will acquire are spaced to schedule carries out resampling；
Data smoothing subelement, the data after resampling are carried out Denoising disposal using data smoothing algorithm.
Feature extraction unit 203, to extract mouse running fix target signature for mouse action sample.Wherein, institute
Stating mouse running fix target signature can specifically include: feature, characterization movement speed of the characterization movement routine planning with adjustment
Feature, the characterization of planning and adjustment move closer to the feature of the mobile adjustment and control of object procedure.
Character analysis module 204, the personality model based on mouse running fix target signature trained in advance to foundation,
The mouse running fix Target Signature Analysis user's personality obtained based on feature extraction unit 203；Wherein, mobile fixed based on mouse
The training method apparatus structure of the personality model of position target signature is as shown in Figure 4.
Referring to Fig. 3, the present invention is based on the personality model training side of mouse running fix target signature is as follows:
Step 301: data prediction being carried out to several mouse action data and obtains several mouse action samples.
It is possible, firstly, to get several mouse action data, such as character test webpage is established, web page contents are property
Lattice test questionnaire, invite several users to move and click on by operation mouse and the answer completion personality test of each problem is selected to ask
Volume can be recorded the mouse action data of user during user answers a questionnaire by JavaScript, complete to ask in user
After volume is answered, the personality that can recorde user is used for subsequent model learning as the label of mouse action data.
Data prediction may include data resampling and data smoothing.
Wherein, data resampling is spaced progress resampling to the mouse action data that will acquire to schedule, this
The predetermined time interval of inventive embodiments is 8ms.Wherein, data smoothing is using data smoothing algorithm by the data after resampling
Denoising disposal is carried out, the data smoothing algorithm that embodiment uses is sliding averaging method.Those skilled in the art can be to data
The predetermined time interval and data smoothing algorithm of resampling separately select, and can also skip data resampling steps directly to acquisition
Mouse action data using data smoothing algorithm carry out Denoising disposal, do not influence realization of the invention.
User's mouse action data that the embodiment of the present invention obtains are shown in Table 1.
The personality measure that the present embodiment is chosen includes 4 character dimensions, and every character dimension includes that two kinds of personality are inclined
To respectively corresponding are as follows: export-oriented-internally-oriented, Intuition-modality of sensation, thoughtful type-emotion type, judgement type-sensing type.
In the present embodiment, it is available corresponding in 4 character dimensions to user after user's completion questionnaire is answered
It is biased to, the label by the deviation of this 4 character dimension as the mouse action data of corresponding user, the mouse of each user
The label of corresponding 4 dimensions of operation data.
Step 302: for several mouse action samples of record, analyzing user's operation mouse control computer screen glazing
Mark is mobile with dynamic change during the close and position that makes it: user's dynamic change movement routine is planned and is adjusted
Whole, movement speed planning and the mobile adjustment and control for adjusting, moving closer to object procedure；It is special to extract mouse running fix target
Sign describes the motion planning control model.The feature that the present embodiment extracts, the example of reference table 2-1,2-2,2-3, mouse are mobile
Positioning target signature example includes the feature for characterizing the planning of mouse movement routine and adjustment, the planning of characterization mouse movement speed and adjusts
Whole feature, and characterization mouse move closer to the feature of the mobile adjustment and control of object procedure.
Step 303: feature selection approach is used, such as minimal redundancy maximum correlation mRMR algorithm, to mouse running fix
Target signature carries out dimensionality reduction.Embodiment drops primitive character while selecting validity feature using feature selection approach
It ties up, ensure that the minimum between feature is superfluous using the characteristic set that minimal redundancy maximum correlation mRMR algorithms selection goes out in embodiment
The maximum correlation of remaining property and feature and class label, in specific implementation, those skilled in the art can choose other spies
It levies selection method or dimension-reduction algorithm realizes selection and the dimensionality reduction of validity feature, do not influence realization of the invention.
Step 304: use sorting algorithm, as support vector machines to the mouse running fix target signature after dimensionality reduction into
Row classification, or regression algorithm is used, as support vector regression SVR returns the mouse running fix target signature after dimensionality reduction
Return, obtains the corresponding personality model based on mouse running fix target signature of different characters characteristic.
The personality measure that the embodiment of the present invention is chosen includes 4 character dimensions, and every character dimension includes two kinds of property
Lattice are biased to, and are respectively corresponded are as follows: export-oriented-internally-oriented, Intuition-modality of sensation, thoughtful type-emotion type, judgement type-sensing type.?
In embodiment, the personality model based on mouse running fix target signature may include 4 two classification SVM classifiers, training 4
Two classification SVM classifiers correspond respectively to 4 character dimensions for inferring that user is biased in the personality of every character dimension.For
Each two classification SVM classifier, specific training step may include:
The mouse running fix target signature after dimensionality reduction is projected into higher dimensional space using mapping function, is adopted in the present embodiment
Mapping function is RBF function；Based on the higher dimensional space mouse running fix target signature training SVM classifier after projection；
Training result is stored as personality disaggregated model.In practical applications, those skilled in the art can choose other mapping functions
Or other machines learning method training personality model, realization of the invention is not influenced.
With reference to Fig. 4, the personality model training method based on mouse running fix target signature provided with aforementioned present invention
Corresponding, the personality model training apparatus the present invention is based on mouse running fix target signature includes with lower unit:
Personality model training pretreatment unit 401: if being obtained to carry out data prediction to several mouse action data
Dry mouse action sample.Wherein, personality model training pretreatment unit 401, can specifically include:
Data resampling sub-units, the mouse action data will acquire are spaced to schedule carries out resampling；
Data smoothing subelement, the data after resampling are carried out Denoising disposal using data smoothing algorithm.
Personality model training feature extraction unit 402: to extract mouse running fix for several mouse action samples
Target signature.Wherein, mouse running fix target signature can specifically include: the feature of the planning of characterization movement routine and adjustment,
Feature, the characterization of the planning of characterization movement speed and adjustment move closer to the feature of the mobile adjustment and control of object procedure.
Feature Dimension Reduction unit 403: right such as minimal redundancy maximum correlation mRMR algorithm to use feature selection approach
The mouse running fix target signature carries out dimensionality reduction.
Model learning unit 404: to use sorting algorithm, as support vector machines are mobile to the mouse after dimensionality reduction fixed
Position target signature is classified, or uses regression algorithm, if support vector regression SVR is to the mouse running fix mesh after dimensionality reduction
Mark feature is returned, and the corresponding personality model based on mouse running fix target signature of different characters characteristic is obtained.
Wherein, model learning unit 404 can specifically include projection subelement, classification subelement, return subelement, storage
Subelement.Projection subelement is to use mapping function that the mouse running fix target signature after dimensionality reduction is projected to higher-dimension sky
Between；Classify subelement such as to support based on the higher dimensional space mouse running fix target signature training classification method after projection
Vector machine SVM；Return subelement to based on after projection higher dimensional space motion control feature training homing method, such as support to
Amount returns SVR；Storing sub-units are to be stored as personality disaggregated model or personality regression model for training result.
Claims (10)
1. a kind of online character analysis method based on motion planning controlling feature, which comprises the following steps:
Step 1: obtaining and recording the motion planning control data of active user；
Step 2: carrying out data prediction to motion planning control data, the sample action of user is obtained；
Step 3: the motion planning mode in analysis sample action, extracts the motion planning controlling feature in sample action；
Step 4: carrying out character analysis to motion planning controlling feature by the personality model pre-set, wherein based on fortune
The personality model of dynamic planning control feature is the different characters property sort device obtained by the training of several sample actions or recurrence
Device.
2. online character analysis method according to claim 1, it is characterised in that: the step one by computer,
Mobile terminal or wearable device obtain and record the motion planning control data of active user；Obtaining motion planning control
When data, gap of the user based on current motion state and target state is planned and adjusts current motion state to reach
Target state；The motion planning mode for obtaining motion planning control data includes mouse running fix target motion planning mould
Formula, keyboard input predetermined sequence motion planning mode, touch screen input prearranged gesture motion planning mode, everyday actions movement rule
The mode of drawing；
The mouse running fix target motion planning mode are as follows: user is moved using cursor on mouse control computer screen
Move and make it position when, view-based access control model perception computer comparison screen on present cursor position and target position
Information, it is dynamic change that operation mouse, which moves closer to the cursor controlled on screen and reach the process of target position, includes
The mobile adjustment and control for adjustment, moving closer to object procedure are planned in movement routine planning with adjustment, movement speed；
The keyboard inputs predetermined sequence motion planning mode are as follows: user completes predetermined keystroke sequence input using keyboard
When target, based on the mobile keystroke of the relative position information of finger and keyboard control finger, the keystroke section that reflection finger movement is coordinated
It plays；
The touch screen inputs prearranged gesture motion planning mode are as follows: user completes prearranged gesture input using touch screen
When target, the display information change based on intelligent mobile terminal screen, control finger is slided to complete the mistake of corresponding gesture input
Journey is dynamic change, is adjusted and control, sliding speed adjustment and control, sliding path adjustment and control comprising sliding pressure；
The prearranged gesture includes pull-up, drop-down, left cunning, right cunning, zoom-in and zoom-out；
The everyday actions motion planning mode are as follows: user is wearing mobile terminal or wearable device completion everyday actions
When, habit is controlled by the everyday actions that the data that motion sensor records obtain user.
3. online character analysis method according to claim 2, it is characterised in that: the sample action includes mouse behaviour
Make the everyday actions sample of sample, keyboard key stroke sample, touch screen sliding sample, motion sensor record；The movement rule
Drawing controlling feature includes mouse running fix target motion planning controlling feature, keyboard input predetermined sequence motion planning control spy
Sign, touch screen input prearranged gesture motion planning controlling feature and everyday actions motion planning controlling feature；
The shifting that interactive device, user when motion planning control data pass through user's operation computer or intelligent mobile terminal are worn
The motion sensor measurement of dynamic terminal and wearable device obtains；Wherein, user's operation computer or when intelligent mobile terminal
Interactive device include mouse, keyboard, touch screen, interbehavior data include mouse action data, keyboard key stroke data and touching
Screen sliding data are touched, everyday actions data are recorded by the motion sensor of mobile terminal or wearable device.
4. online character analysis method according to claim 3, it is characterised in that: the mouse running fix target fortune
Dynamic planning control feature includes the feature for characterizing the planning of mouse movement routine and adjustment, the planning of characterization mouse movement speed and adjustment
Feature, characterization mouse move closer to object procedure mobile adjustment and control feature；
1) planning of characterization mouse movement routine and the feature of adjustment include:
During mobile mouse, when cursor position deviates central axes farther out, user, which adjusts move angle, returns cursor for description
Move angle-distance of distance of coordinate when returning central axes to target location coordinate coordinates feature；
It describes traveling time and present cursor position and coordinates spy to traveling time-distance of relationship between the distance of target position
Sign；
Mobile curvature-time coordination feature of mobile bending degree dynamic adjustment and traveling time relationship between the two is described；
The mobile curvature statistic of mobile bending degree globality variation is described；
The mobile deviation central axes statistic of the mobile offset central axes globality variation of description；
The moving displacement of mobile efficiency and the ratio of distance are described；
2) planning of characterization mouse movement speed and the feature of adjustment include:
Description is mobile accelerate and slow down dynamic adjustment and present cursor position to target position distance relationship between the two
Movement speed-distance coordinates feature；
Description is mobile accelerate and slow down dynamic adjustment and traveling time relationship between the two movement speed-time coordination it is special
Sign；
The movement speed statistic of movement speed globality variation is described；
Movement speed is described with the translational acceleration statistic of time change；
3) feature for mobile adjustment and the control that characterization mouse moves closer to object procedure includes:
The ratio close to target traveling time and whole section of traveling time close to target adjustment time span is described；
Describe to change close to movement speed globality in the moving process of target close to target velocity statistic；
Description close in the moving process of target mobile curvature globality variation close to target curvature statistic；
The ratio close to target moving displacement and distance of the mobile efficiency close in the moving process of target is described；
Residence time of the mobile finish time of mouse task switching efficiency to the click moment between the two is described；
Whether moving for description mistake adjustment is back adjusted more than target to reach target.
5. online character analysis method according to claim 3, it is characterised in that: the keyboard input predetermined sequence fortune
Dynamic planning control feature includes the feature for the keystroke rhythm that characterization reflection finger movement is coordinated；
The feature of the keystroke rhythm that the described characterization reflection finger movement is coordinated includes describing the key-efficient of each key to hit
Key residence time statistic, the keystroke flight time statistic of the key-efficient of the adjacent key twice of description, description it is given two
Double character group keystroke flight time statistics of the key-efficient of a letter or the corresponding key pair of number.
6. online character analysis method according to claim 3, it is characterised in that:
The touch screen input prearranged gesture motion planning controlling feature includes the feature characterized sliding pressure adjustment with control, table
Levy the feature of sliding speed adjustment and control, the feature of characterization sliding path adjustment and control；
1) feature of the characterization sliding pressure adjustment and control includes:
Describe sliding pressure dynamic adjustment with sliding distance between the two relationship sliding pressure-distance coordination feature；
Sliding pressure-time coordination feature of sliding pressure dynamic adjustment and sliding time relationship between the two is described；
The sliding pressure statistic of sliding pressure globality variation is described；
2) feature of the characterization sliding speed adjustment and control includes:
Describe sliding speed dynamic adjustment with sliding distance between the two relationship sliding speed-distance coordination feature；
Sliding speed-time coordination feature of sliding speed dynamic adjustment and sliding time relationship between the two is described；
The sliding speed statistic of sliding speed globality variation is described；
The sliding acceleration statistic of description sliding acceleration globality variation；
3) feature of the characterization sliding path adjustment and control includes:
The slip distance of sliding length is described；
Describe glide direction dynamic adjustment with sliding distance between the two relationship glide direction-distance coordination feature；
Glide direction-time coordination feature of glide direction dynamic adjustment and sliding time relationship between the two is described；
Sliding curvature-time coordination feature of description sliding bending degree dynamic adjustment and sliding time relationship between the two；
The sliding curvature statistic of description sliding bending degree.
7. online character analysis method according to claim 3, it is characterised in that: the everyday actions motion planning control
Feature includes characterizing the feature of the travelling control habit in user's walking process；The travelling control characterized in user's walking process is practised
Used feature include: user is described to walk the feature of rhythm, including walking cycle, acceleration of averagely walking, the average speed of travel,
Average step length；When describing user's walking, the body bobbing feature stopped over body bobbing rule during foot-up；User one is described
Walking acceleration-travel time of the adjustment of acceleration dynamic and travel time relationship between the two is coordinated special in a walking cycle
Sign.
8. online character analysis method according to claim 1, it is characterised in that: the step four is based on motion planning
The personality model of controlling feature is carried out as follows foundation:
4-1, several sample actions are obtained to several motion plannings control data progress data prediction；It is dynamic to analyze several
Make the motion planning mode in sample, and for the motion planning controlling feature in the motion planning schema extraction sample action；
4-2, dimensionality reduction is carried out to the motion planning controlling feature by feature selecting algorithm；
4-4, classified using sorting algorithm to the motion planning controlling feature after dimensionality reduction, or using regression algorithm to dimensionality reduction
Motion planning controlling feature afterwards is returned, and the personality model that different characters characteristic corresponds to motion planning controlling feature is obtained.
9. a kind of device for realizing the online character analysis method described in claim 1 based on motion planning controlling feature, special
Sign is: including the data capture unit for obtaining and recording active user's motion planning control data, for record
Motion planning control data carry out data prediction and obtain the pretreatment unit of user action sample, for analyzing in sample action
Motion planning mode and for the feature extraction of the motion planning controlling feature in the motion planning schema extraction sample action
Unit, and the character analysis for carrying out character analysis to motion planning controlling feature according to the personality model pre-set
Unit.
10. device according to claim 9, it is characterised in that:
The pretreatment unit can control several motion plannings data progress data prediction and obtain several sample actions,
Feature extraction unit can analyze the control of the motion planning in the motion planning schema extraction sample action in several sample actions
Feature carries out dimensionality reduction to motion planning controlling feature by Feature Dimension Reduction unit, uses sorting algorithm by model learning unit
Classify to the motion planning controlling feature after dimensionality reduction, or using regression algorithm to the motion planning controlling feature after dimensionality reduction
It is returned, obtains the personality model that different characters characteristic corresponds to motion planning controlling feature；
The feature extraction unit includes that mouse running fix target motion planning controlling feature extraction unit, keyboard input are pre-
Sequencing column motion planning controlling feature extraction unit, touch screen input prearranged gesture motion planning controlling feature extraction unit and
Everyday actions motion planning controlling feature extraction unit；Wherein, the mouse running fix target motion planning controlling feature
Extracting unit used for extracting characterizes the feature of the planning of mouse movement routine and adjustment, the planning of characterization mouse movement speed and adjustment
Feature, characterization mouse move closer to the feature of the mobile adjustment and control of object procedure；The keyboard input predetermined sequence fortune
Feature of the dynamic planning control feature extraction unit to extract the keystroke rhythm that characterization reflection finger movement is coordinated；The touch
Feature, the table of screen input prearranged gesture motion planning controlling feature the extracting unit used for extracting adjustment of characterization sliding pressure and control
Levy the feature of sliding speed adjustment and control, the feature of characterization sliding path adjustment and control；The everyday actions move rule
Draw the feature of the travelling control habit in controlling feature extracting unit used for extracting characterization user's walking process；
The model learning unit includes projection subelement, classification subelement, returns subelement and storing sub-units；Projection
Unit is to use mapping function by the motion control Projection Character after dimensionality reduction to higher dimensional space, and subelement of classifying is to based on throwing
The higher dimensional space motion control feature training classification method of movie queen returns subelement to based on the higher dimensional space movement after projection
Controlling feature trains homing method, and training result to be stored as the disaggregated model of different characters characteristic or returned by storing sub-units
Return model.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910329694.6A CN110110766B (en) | 2019-04-23 | 2019-04-23 | Online character analysis method and device based on motion planning control characteristics |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910329694.6A CN110110766B (en) | 2019-04-23 | 2019-04-23 | Online character analysis method and device based on motion planning control characteristics |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN110110766A true CN110110766A (en) | 2019-08-09 |
| CN110110766B CN110110766B (en) | 2021-08-13 |
Family
ID=67486364
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910329694.6A Active CN110110766B (en) | 2019-04-23 | 2019-04-23 | Online character analysis method and device based on motion planning control characteristics |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110110766B (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116739814A (en) * | 2023-04-23 | 2023-09-12 | 广州市疾病预防控制中心(广州市卫生检验中心、广州市食品安全风险监测与评估中心、广州医科大学公共卫生研究院) | Method for preventing disease transmission and social platform |
Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101833619A (en) * | 2010-04-29 | 2010-09-15 | 西安交通大学 | Method for judging identity based on keyboard-mouse crossed certification |
| CN101833626A (en) * | 2010-05-19 | 2010-09-15 | 西安交通大学 | Method for verifying computer user identity based on keystroke scrambling characteristic |
| CN103530540A (en) * | 2013-09-27 | 2014-01-22 | 西安交通大学 | User identity attribute detection method based on man-machine interaction behavior characteristics |
| CN106940805A (en) * | 2017-03-06 | 2017-07-11 | 江南大学 | A kind of group behavior analysis method based on mobile phone sensor |
| CN107688827A (en) * | 2017-08-24 | 2018-02-13 | 西安交通大学 | A kind of user identity attribute forecast method based on user's daily behavior feature |
| CN108182489A (en) * | 2017-12-25 | 2018-06-19 | 浙江工业大学 | Method is recommended in a kind of individualized learning based on on-line study behavioural analysis |
| CN109165692A (en) * | 2018-09-06 | 2019-01-08 | 中国矿业大学 | A kind of user's personality prediction meanss and method based on Weakly supervised study |
-
2019
- 2019-04-23 CN CN201910329694.6A patent/CN110110766B/en active Active
Patent Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101833619A (en) * | 2010-04-29 | 2010-09-15 | 西安交通大学 | Method for judging identity based on keyboard-mouse crossed certification |
| CN101833626A (en) * | 2010-05-19 | 2010-09-15 | 西安交通大学 | Method for verifying computer user identity based on keystroke scrambling characteristic |
| CN103530540A (en) * | 2013-09-27 | 2014-01-22 | 西安交通大学 | User identity attribute detection method based on man-machine interaction behavior characteristics |
| CN106940805A (en) * | 2017-03-06 | 2017-07-11 | 江南大学 | A kind of group behavior analysis method based on mobile phone sensor |
| CN107688827A (en) * | 2017-08-24 | 2018-02-13 | 西安交通大学 | A kind of user identity attribute forecast method based on user's daily behavior feature |
| CN108182489A (en) * | 2017-12-25 | 2018-06-19 | 浙江工业大学 | Method is recommended in a kind of individualized learning based on on-line study behavioural analysis |
| CN109165692A (en) * | 2018-09-06 | 2019-01-08 | 中国矿业大学 | A kind of user's personality prediction meanss and method based on Weakly supervised study |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116739814A (en) * | 2023-04-23 | 2023-09-12 | 广州市疾病预防控制中心(广州市卫生检验中心、广州市食品安全风险监测与评估中心、广州医科大学公共卫生研究院) | Method for preventing disease transmission and social platform |
| CN116739814B (en) * | 2023-04-23 | 2024-05-14 | 广州市疾病预防控制中心(广州市卫生检验中心、广州市食品安全风险监测与评估中心、广州医科大学公共卫生研究院) | Method for preventing disease transmission and social platform |
Also Published As
| Publication number | Publication date |
|---|---|
| CN110110766B (en) | 2021-08-13 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Escalera et al. | Challenges in multi-modal gesture recognition | |
| Mohamad Nezami et al. | Automatic recognition of student engagement using deep learning and facial expression | |
| Huynh et al. | Detection of driver drowsiness using 3D deep neural network and semi-supervised gradient boosting machine | |
| US9690982B2 (en) | Identifying gestures or movements using a feature matrix that was compressed/collapsed using principal joint variable analysis and thresholds | |
| Liu et al. | Recent progress on tactile object recognition | |
| Gao et al. | What does touch tell us about emotions in touchscreen-based gameplay? | |
| Frommel et al. | Towards emotion-based adaptive games: Emotion recognition via input and performance features | |
| Banjarey et al. | A survey on human activity recognition using sensors and deep learning methods | |
| Shah et al. | Towards affective touch interaction: predicting mobile user emotion from finger strokes | |
| CN104778600A (en) | Incentive mechanisms for user interaction and content consumption | |
| Wang et al. | Clustering-based emotion recognition micro-service cloud framework for mobile computing | |
| Russel et al. | Gender discrimination, age group classification and carried object recognition from gait energy image using fusion of parallel convolutional neural network | |
| Piana et al. | Automated analysis of non-verbal expressive gesture | |
| Wei et al. | Performance monitoring and evaluation in dance teaching with mobile sensing technology | |
| Gamberini et al. | How natural is a natural interface? An evaluation procedure based on action breakdowns | |
| Thiam et al. | A temporal dependency based multi-modal active learning approach for audiovisual event detection | |
| Wang et al. | Low personality-sensitive feature learning for radar-based gesture recognition | |
| CN110110766A (en) | A kind of online character analysis method and device based on motion planning controlling feature | |
| Kandemir et al. | Learning relevance from natural eye movements in pervasive interfaces | |
| Ahmed et al. | Two-layer feature selection algorithm for recognizing human emotions from 3d motion analysis | |
| Zhang | Human–Computer Interactive Gesture Feature Capture and Recognition in Virtual Reality | |
| Kirchhof et al. | Chances of interpretable transfer learning for human activity recognition in warehousing | |
| Calapatia et al. | Using body posture, movement and human activity to predict emotions | |
| Li et al. | Controlling non-touch screens as touch screens using Airpen, a writing tool with in-air gesturing mode | |
| Cecotti et al. | Hand-drawn symbol recognition in immersive virtual reality using deep extreme learning machines |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |